[llvm] [LAA] Use MaxStride instead of CommonStride to calculate MaxVF (PR #98142)

[Florian Hahn via llvm-commits]

================
@@ -0,0 +1,156 @@
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 5
+; RUN: opt --disable-output -mtriple=x86_64 --passes="print<access-info>" %s 2>&1 | FileCheck %s
+
+ at a = dso_local local_unnamed_addr global [65536 x float] zeroinitializer, align 16
+
+; Generated from the following C code:
+; #define LEN 256 * 256
+; float a[LEN];
+;
+; void different_strides() {
+;   for (int i = 0; i < LEN - 1024 - 255; i++) {
+;   #pragma clang loop interleave(disable)
+;   #pragma clang loop unroll(disable)
+;     for (int j = 0; j < 256; j++)
+;       a[i + j + 1024] += a[j * 4 + i];
+;   }
+; }
+; The load and store have different strides(4 and 16 bytes respectively) but the store
+; is always at safe positive distance away from the load, thus BackwardVectorizable
+define dso_local void @different_strides_backward_vectorizable() local_unnamed_addr {
----------------
fhahn wrote:

```suggestion
define void @different_strides_backward_vectorizable() {
```

https://github.com/llvm/llvm-project/pull/98142